Mill Computing, Inc. › Forums › Announcements › Events › What is your roadmap for 2023?
- NarrateurDuChaosParticipantJanuary 6, 2023 at 6:52 amPost count: 22#3779 |
I’m really interested in knowing how the Mill project is progressing, but the process so far is opaque. So I thought I’d ask some progress-related questions!
– What do you hope to complete before 2023?
– Are there any milestones you hope to complete that would warrant a public announcement?
– When do you think a simulator will be available for testing? (Six months? A year? Five years?)
– When do you think a first FPGA implementation will be completed?
Hope you all make lots of progress this year!
- Ivan GodardKeymasterJanuary 7, 2023 at 11:42 pmPost count: 679
By far the most important thing in 2023 will be our conversion from a bootstrap to a conventional company with paid salaries. We have hit the limit in what can be done in our current structure, so we expect to seek our next funding round, or affiliation with another company, as negotiated and the investment market permits. The answers to your other questions depend on how that goes. Naturally, we can’t talk about the details.
- DeepBlueParticipantFebruary 6, 2023 at 1:48 amPost count: 5
How about crowdfunding and open sourcing the ISA?
Post RISC-V it will be difficult to find traction
without publishing the ISA.
The real value is not the textual ISA, but rather
the microarchitecture, which one would think is
protected by patents.
The industry is long past the stage when organic
growth was possible. To be successful the Mill
will either have to find one killer app, or to
support a broad application portfolio.
- Ivan GodardKeymasterFebruary 7, 2023 at 7:29 pmPost count: 679
Open sourcing would only happen if we could not find funding otherwise. Crowd-funding $15M+ is a bit of a stretch. We will first try the usual funding paths.
- blakliteParticipantJanuary 21, 2023 at 3:52 pmPost count: 1
As a complete novice the whole idea sounds amazing. Given the recent shift(s) in focus towards more hardware based startups, I really do hope you manage to maintain your vision and identity while looking for further funding, hopefully maintaining your obvious passion for the project.
Please let this be allowed to flourish (or fail) on it’s merits, rather than at the whim of short term fiscal whims.
Wishing everyone the best for a stronger 2023 and beyond.
- Ivan GodardKeymasterJanuary 29, 2023 at 5:09 amPost count: 679
QEMU is one of the many things on our wish list that are waiting for our conversion from the bootstrap business model. The interpreter will be non-trivial; it’s not set up for wide ISAs and interleaved execution.
- DeepBlueParticipantJanuary 29, 2023 at 3:41 pmPost count: 5
It does not sound plausible, let alone reasonable, that a
Mill plug-in for QEMU could require a prohibitive level of
effort. IMHO this is a junior year CS semester project for
one person. It is quite likely one or more of Mill’s many
fans would be happy to volunteer.
The argument QEMU is not “set up” for wide ISAs sounds
questionable. First, QEMU is open source and appopriate
support can be added if and where missing. Second, QEMU
has already been used to model and simulate wide ISAs
and interleaved execution by various projects in the
computer industry. One wonders why the Mill project
did not start with a QEMU emulator. It would have
saved time and effort versus building a complete
simulation environment from scratch.
- This reply was modified 4 months, 1 week ago by DeepBlue.
- c3dParticipantFebruary 4, 2023 at 5:11 amPost count: 6
@deepblue I agree with Ivan that there are things that are quite different that require brand new infrastructure to emulate the Mill correctly. For one thing, all the wide ISAs emulated so far are not really in the mainstream qemu support. Support for the most mainstream of them, Itanium, was dropped back in 2017. All the existing platforms also assume a rather classical register file, not a belt with configurable size and so on. So yes, this is going to be a lot of work in my opinion, not a student’s work for one year.
- Ivan GodardKeymasterFebruary 4, 2023 at 7:56 amPost count: 679
- DeepBlueParticipantMarch 1, 2023 at 10:25 pmPost count: 5
“@deepblue I agree with Ivan that there are things that are
quite different that require brand new infrastructure to
emulate the Mill correctly. For one thing, all the wide
ISAs emulated so far are not really in the mainstream
qemu support. Support for the most mainstream of them,
Itanium, was dropped back in 2017. All the existing
platforms also assume a rather classical register file,
not a belt with configurable size and so on. So yes,
this is going to be a lot of work in my opinion, not
a student’s work for”.
The fact you are referring to the “official” list of
supported of ISAs supported by QEMU opens the question
if you have actually ever worked directly with QEMU code.
Did you pick this up from the Wikipedia? Keep in mind
many projects used QEMU internally as a scaffold and
never released any information to the general public.
FWIW I modified QEMU to emulate a vector architecture
many years ago. It was a piece of cake. QEMU code is
quite modular — big and small blocks can be swapped
in and out in pretty straightforward fashion.
- Ivan GodardKeymasterMarch 3, 2023 at 7:01 pmPost count: 679
I personally know next to nothing about QEMU, having never used it. I did write our present sim.
In my ignorance I anticipate the most trouble with representing in-flight values. Because Mill explicitly separates instruction initiation from instruction retire, a sim must model instructions that are in-flight in the pipeline, and merge the effects of separate instructions that emerge from the pipes at the same time despite having been initiated at different times. Due to phasing, the init and retire can be in the same bundle cycle, with other execution in the middle. In particular, there can be control flow during the in-flight period.
You can’t just start at an address and assume that everything before is in the belt/memory. There may be an in-flight multiply that will drop to the belt in two cycles, completely unannounced, with a whole function executed since the init. And you can’t just snapshot at every basic block either – in-flights can carry over the branches.
I’m not saying it’s impossible. I’m saying that it will be difficult and we presently have neither the expertise nor the money to take it on. @QEMU experts – there’s a challenge here available if you’d like to join the Mill effort.
- WorBluxParticipantMarch 26, 2023 at 8:53 amPost count: 2
>@ Ivan “I anticipate the most trouble with representing in-flight values. Because Mill explicitly separates instruction initiation from instruction retire, a sim must model instructions that are in-flight in the pipeline…”
Right, that does sound tricky. But does the end user need a high-performace Sim/emulator for every conASM? If you just want to port genASM code, then building a QUEMU module as a distinct target for the specializer makes sense to me.
- Ivan GodardKeymasterMarch 26, 2023 at 4:09 pmPost count: 679
One could get the specializer to produce a single-instruction sequence without pipeline overlap. That could be QUEMU emulated, but we have a sim already. The other argument for QUEMU is that it can produce translations that are close to native performance on an alien architecture, saving the cost/effort of a port/retarget. But applied to the Mill that would give close to the performance of a single-instruction sequence with no parallelism, or truly awful in other words.
No doubt someone will someday try to extend QUEMU itself to be able to handle Mill-like ISAs. But not us today on our very limited nickels.
- DeepBlueParticipantMarch 1, 2023 at 10:14 pmPost count: 5
Re “Crowd-funding $15M+ is a bit of a stretch”:
Plenty of projects and ventures have raised a
lot more than $15M. Check this list, it starts
just under $60M:
- jakanaradaParticipantApril 4, 2023 at 4:00 amPost count: 1
Hi there, from a long time listener (and first time caller) of the project. Just leaving an observation tag that occurred in the moment (April/2023). It’s easy for outsiders to offer opinions without knowing the reality of the situation within the project, but assuming crowdfunding is not an option, and project has not attempted this, maybe try: Nomura Japan, AgBank China and UBS Switzerland (in that order). After all, there is a global arms race going on in computing that is only just beginning to heat-up with CBDCs & CBDC Monitoring/Audit AI and Blackrock’s proclamation in 2022/2023 that ‘Tokens Are “The Next Generation For Markets”’ – not cryptocurrencies but the underlying smart contract facilitating token systems.
The trigger for the observation came from a scene from the 1997 movie, Contact.
To paraphrase the scene:
“Look, all I’m asking is for you to just have the tiniest bit of vision. You know, to just sit back for one minute and look at the big picture. To take a chance on something that just might end up being the most profoundly impactful moment for humanity, for the history… of technology.”
Usual disclaimers apply: This is not legal advice and this is not financial advice. This post is a random off-the-cuff observation that occurred in the moment.
- Ivan GodardKeymasterApril 4, 2023 at 11:12 pmPost count: 679
Thank you for the suggestions. What would be even more useful would be an introduction to a person within those institutions. Anyone?
You must be logged in to reply to this topic.